Should PMs Have Codebase Access Now That AI Coding Tools Exist?
I have a pretty strong opinion on this:
Product managers should have read access to the codebase of the product they work on.
Not because PMs should suddenly start reviewing PRs or tell engineers how to build things. That would obviously be the wrong takeaway.
But with tools like Claude Code, Cursor, Codex, etc., this becomes a much more practical question than it was a few years ago.
A PM no longer needs to understand every implementation detail to get value from the codebase. They can search through it, ask questions about flows, understand where certain product behavior lives, and get a better feeling for how the system is actually structured.
And honestly, a lot of “product reality” lives in the codebase.
Things like permissions, validation logic, edge cases, API contracts, feature flags, background jobs, integrations, weird legacy assumptions, etc.
For me, this is not about taking ownership away from engineering. It is more about being able to ask better questions, understand constraints earlier, and not treat every technical topic as a black box.
Of course there is a line. PMs should not overstep, jump to implementation conclusions, or use AI coding tools to bypass conversations with engineers.
But I do think the rise of tools like Claude Code, Cursor and Codex changes the bar for what technical literacy can look like in product management.
Curious how others see this.
Should PMs have read access to the codebase?
And do AI coding tools make that more reasonable, or more dangerous?